Reactant-Induced\nDynamic Active Sites on Cu Catalysts\nduring the Water–Gas Shift Reaction

Adsorbates can trigger surface reconstruction on metal\nsurfaces,\na common yet highly important phenomenon in heterogeneous catalysis\nthat has not been fully explored. Here, we develop a reliable Cu–C–O\nmachine learning force field (MLFF) with ab initio accuracy, providing\ninsights into the reconstruction mechanism and distribution of active\nsites on the Cu surface under a CO atmosphere through state-of-the-art\ndeep potential molecular dynamics (DPMD). Combining statistical cluster\nanalysis with microkinetic modeling, we establish a strategy to quantitatively\nassess the turnover frequency (TOF) of catalyst surfaces during the\ndynamic catalytic process. Our findings reveal that edge Cu atoms\nundergo rearrangement, ejection, diffusion, and aggregation under\na CO atmosphere, leading to the formation of cluster active sites.\nThese small clusters in dynamic equilibrium are identified as the\norigin of the high catalytic activity of Cu-based catalysts for a\nlow-temperature water–gas shift reaction (WGSR). This work\nnot only elucidates intrinsic activity in metal catalysis and the\ndynamic catalysis theory but also offers valuable insights for computational\ncatalysis methods to identify effective catalysts for practical applications.